[clang] [BoundsSafety][Sema] Allow counted_by and counted_by_or_null on pointers where the pointee type is incomplete but potentially completable (PR #106321)

[via cfe-commits]

================
@@ -8207,6 +8207,14 @@ ExprResult InitializationSequence::Perform(Sema &S,
             Kind.getRange().getEnd());
       } else {
         CurInit = new (S.Context) ImplicitValueInitExpr(Step->Type);
+        // Note the return value isn't used to return early
+        // to preserve the AST as best as possible even though an error
+        // might have occurred. For struct initialization it also allows
+        // all field assignments to be checked rather than bailing on the
+        // first error.
----------------
Sirraide wrote:

> It would be superfluous if not for the part about struct initialization. To me that is really not very obvious.

That’s fair. I might just be to used to this as a pattern. Commenting *why* we don’t just throw the entire thing out of the window here and still try to keep operating on it even if this fails is probably fine.

> I've changed the comments to just be.
> 
> ```
>         // Note the return value isn't used to return a ExprError() when
>         // initialization fails . For struct initialization allows all field
>         // assignments to be checked rather than bailing on the first error.
> ```

I’d honestly shorten it even more to just:
```
// Don't return here so we can still check field assingments.
```
because e.g. the fact that ‘the return value isn't used to return a ExprError() when initialization fails’ is apparent from the fact that... there is no `return` statement here. ;Þ


https://github.com/llvm/llvm-project/pull/106321